The biochemical rationale for AZT therapy in AIDS is twofold. On the one hand, AZT-triphosphate is a substrate for HIV-I reverse transcriptase and incorporation of this dideoxynucleoside (ddN) chain terminator into HIV-1 DNA inhibits its replication. On the other, it is neither a substrate nor an inhibitor of DNA polymerase alpha, leaving host nuclear DNA replication unimpaired. Several other ddN analogs behave similarly. Nevertheless, many ddn's are toxic, inducing anemia (AZT), peripheral neuropathy (ddC, d4T, ddI) and pancreatitis (ddI). Our hypothesis accounts for at least some of the toxicity as follows: ddn-triphosphates are good substrates for DNA polymerase gamma leading to, in turn: their incorporation into mitochondrial DNA; inhibition of mtDNA replication by chain termination; gradual decrease in mitochondrial number; decrease in rate of ATP synthesis; decrease in cellular ATP level. Evidence for this hypothesis was obtained using several systems, namely: DNA polyrase gamma; isolated mitochondria; differentiating Friend erythroleukemic cells which synthesize hemoglobin, a model for human bone marrow; human bone marrow itself. The results obtained with twelve ddN analogs, show a close and probably causal association between ddn-inhibition of cell growth and of mtDNA replication. Long term objectives are: to obtain conclusive evidence for our hypothesis; to find a way of decreasing ddN toxicity; using the above cell lines to identify existing or new ddN analogs having higher cytotherapeutic indices than do current drugs. Specific aims are: to show that Friend cells whose growth is slowed by AZT possess mitochondria deficient in both synthesis and level of ATP; using the analogs which induce peripheral neuropathy, to repeat these Friend cell experiments on the PC12 neuronal cell, including studying their effect on PC12 mtDNA replication; to attempt to decrease ddN toxicity, with substances capable of stimulating ATP production using the Friend and PC12 cells; to study intensively two analogs which our results show to be promising, namely anhydro-AZT and anhydro-N3,-UdR; to test twelve additional new anti-HIV-1 ddN analogs in our possession in the hope that the total of twenty four drugs we have studied will provide a clue to the structure/toxicity relationship of ddN analogs.